Abstract

Data assimilation (DA) refers to a family of methods used to synchronize a dynamical model to sparse or noisy measurements of model states. In this paper, we propose a wearable DA platform for neurological research and report our progress in translating a DA computational framework from desktop computation to embedded computation. The unscented Kalman filter (UKF) and a neural mass model (NMM) for sleep-wake regulation are introduced. Next, selection of suitable UKF parameters through MATLAB simulations is described. Finally, four variations of the DA framework are run on an embedded microprocessor in order to find the variation that minimizes computation time while maintaining state reconstruction accuracy. By reducing computational precision of the equation integrator and using a piecewise-linear approximation in place of the tanh function, we increased computational speed by a factor of 3.6 while maintaining a high level of state reconstruction fidelity.

Original languageEnglish (US)
Title of host publicationBioCAS 2019 - Biomedical Circuits and Systems Conference, Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781509006175
DOIs
StatePublished - Oct 2019
Event2019 IEEE Biomedical Circuits and Systems Conference, BioCAS 2019 - Nara, Japan
Duration: Oct 17 2019Oct 19 2019

Publication series

NameBioCAS 2019 - Biomedical Circuits and Systems Conference, Proceedings

Conference

Conference2019 IEEE Biomedical Circuits and Systems Conference, BioCAS 2019
CountryJapan
CityNara
Period10/17/1910/19/19

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence
  • Biomedical Engineering
  • Electrical and Electronic Engineering
  • Instrumentation

Fingerprint Dive into the research topics of 'Toward a Wearable Data Assimilation Platform'. Together they form a unique fingerprint.

Cite this